Document resource
Introduction The UK National Endoscopy Database (NED) APRIQOT trial demonstrated that monthly automated feedback of colonoscopy key performance indicators (KPIs) to endoscopists improved polyp detection within a randomised controlled trial setting. This year-long pilot, funded by the Health Foundation, tests the feasibility and impact of reports delivered on a quarterly basis outside a trial. Now at its mid-point, we aimed to evaluate the impact of the pilot.Methods All procedures performed at enrolled centres were extracted from NED from the pre- (Apr-Jun 2025) and post-intervention (Jul-Oct 2025) periods. For each endoscopist in both periods, median withdrawal time, Buscopan® use, caecal intubation rate (CIR), moderate/severe discomfort rate, polyp detection rate (PDR), proximal PDR, mean number of polyps (MNP) and case-mix adjusted mean number of polyps (aMNP) were calculated and compared using Wilcoxon signed-rank tests. Procedure-level mixed effects models were fitted with endoscopist as a random effect, adjusting for age group, sex and indication for colonoscopy. A random slope for period assessed variation in improvement. Finally, an endoscopist-level linear regression model examined if engagement with feedback reports (sum of individual email and report views and web page visits) predicted improvement in PDR, adjusting for baseline PDR.Results Data from 125,521 procedures performed by 744 endoscopists were analysed. Pairwise comparison showed several KPIs improved after the Introduction of automated feedback. Buscopan use doubled (median 5.01% → 10.20%, p<0.001), and small but significant gains were seen in PDR (increase of 0.60%, p=0.005) and MNP (+3.19 polyps/100 procedures, p=0.022). The improvement in aMNP was non-significant (+2.0 polyps/100 procedures, p=0.096). In the mixed-effects models, post-intervention procedures had 6% higher adjusted odds of polyp detection (aOR 1.06, 95% CI 1.03-1.09, p<0.001) and a 4% higher adjusted rate of polyps detected per procedure (aIRR 1.04, 95% CI 1.03-1.06, p<0.001). Allowing random slopes indicated modest heterogeneity of improvement across endoscopists (slope SD 0.18; intercept-slope corr. -0.34), suggesting greater improvement among lower-baseline performers. Engagement analysis showed higher total interaction with the reports and NED predicted greater PDR improvement after adjusting for baseline (β 0.0011 ± 0.00035, p=0.001). Comparing endoscopists, those who opened at least one report demonstrated a 3.1% improvement in PDR compared to those who didn’t (β 0.031, SE 0.012, p=0.007).Conclusions Despite targeting all endoscopists rather than just those engaged in a trial, and lengthening the gap between reports to quarterly, the automated feedback reports significantly improve polyp detection. As was seen in the original trial, endoscopists with lower baseline performance and those engaging more with feedback saw the greatest gains, highlighting the value of sustained data-driven feedback.